Amazon Nova
Amazon Web Services
Amazon Nova 2 is an AWS-native model suite spanning multimodal reasoning, speech, and embeddings. Nova 2 Lite is fast and operationally convenient inside Bedrock; its independent non-reasoning score is modest, and Nova 2 Pro remains preview-only.
CURRENT MODEL SNAPSHOT
Provider: Amazon Web Services
Current anchor: Nova 2 Lite; Nova 2 Pro remains preview
Lifecycle: Current
Weights: Closed
Reviewed: 21 July 2026
Amazon Nova 2 is a cloud portfolio, not a benchmark-first flagship
Nova 2 Lite is the current generally available reasoning anchor in Amazon Bedrock. It handles text, images, video, and documents, supports a one-million-token context window, can produce up to 65,536 output tokens, and offers extended thinking and tool use. Nova 2 Pro targets more demanding reasoning but remains in preview or early access.
The wider Nova 2 family includes Sonic for speech and Multimodal Embeddings, so its enterprise proposition is architectural consistency across AWS-native modalities rather than a single model winning every intelligence table. Identity, networking, observability, guardrails, and procurement already live in the same control plane for many AWS customers.
The models are proprietary. Portability means portable prompts and evaluation assets, not portable weights.
The independent Lite result favors speed over frontier reasoning
Artificial Analysis reports an Intelligence Index score of 12 for Nova 2 Lite in the evaluated non-reasoning configuration, with about 149 output tokens per second and listed pricing near $0.30 per million input tokens and $2.50 per million output tokens. The reasoning modes can behave differently and must be evaluated separately.
That score makes the positioning clear. Lite is not the default choice for the hardest general reasoning task. It can still be the better enterprise component when low latency, multimodal input, long context, Bedrock governance, and predictable integration outweigh aggregate benchmark rank.
Do not report the non-reasoning score as the result for every Nova 2 Lite thinking mode.
Preview claims for Nova 2 Pro should not be used as production evidence.
AWS-native controls reduce integration effort but increase dependence on one cloud control plane.
The enterprise case: operational convenience inside AWS
Nova is strongest for teams already standardized on AWS that need multimodal processing, speech, embeddings, and reasoning behind Bedrock controls. It is less attractive for organizations seeking open weights, cross-cloud portability, or the highest independently measured reasoning quality.
How we would evaluate it
Test Lite in both relevant reasoning settings on a real Bedrock workflow with Guardrails, private networking, tool calls, and cost attribution. Track completed outcomes, unsupported claims, latency, token use, reviewer time, and operational effort. Keep Pro tests explicitly separate while it is in preview.
Evidence used
Beam AI support status
Under evaluation. No general Beam production-support claim applies across Nova 2 Lite, Pro preview, Sonic, embeddings, regions, and Bedrock configurations.
Use case 1
AWS-native multimodal intake
Process documents, screenshots, and video through Bedrock into a typed case record. Measure extraction accuracy, citations, unsupported inference, latency, cost, and Guardrails behavior.
Use case 2
Long-context knowledge assistant
Use Nova 2 Lite with explicit retrieval and a 1M-window stress test. Plant distant contradictions and measure evidence recall, citations, instruction retention, output cost, and reviewer correction.
Use case 3
Voice workflow with Nova Sonic
Evaluate real-time speech under background noise, interruptions, accents, redaction requirements, and tool calls. Track task completion, latency, transcription errors, unsafe actions, and handoff quality.
Use case 4
Bedrock governance benchmark
Test identity, private networking, encryption, logs, retention, regional controls, Guardrails, quotas, budget alerts, and incident response as part of the model evaluation.
Related LLMs
Curated alternatives to compare before selecting a model family.
FAQs
Frequently Asked Questions
Model selection, deployment, governance, and Beam support questions answered.
What is the current Amazon Nova model?
Nova 2 Lite is the generally available reasoning anchor. Nova 2 Pro targets higher capability but remains preview or early access, while Sonic and Multimodal Embeddings cover speech and retrieval use cases.
How strong is Nova 2 Lite independently?
Artificial Analysis reports a score of 12 for the evaluated non-reasoning configuration, about 149 output tokens per second, and listed pricing of roughly $0.30 input and $2.50 output per million tokens. Reasoning modes require separate tests.
Is Amazon Nova open-weight?
No. Nova is a proprietary AWS model family delivered through Amazon services such as Bedrock. Plan for contractual and technical AWS dependence.
Why choose Nova over a higher-scoring model?
For AWS customers, Nova can reduce integration and governance effort across multimodal, speech, embeddings, identity, networking, logs, and procurement. The right comparison is total accepted-outcome cost, not benchmark rank alone.
Does Beam support Amazon Nova in production?
Beam support is currently Under evaluation. No blanket production-support claim applies across all Nova models, regions, or Bedrock configurations.